WO2013056733A1 - Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique - Google Patents

Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique Download PDF

Info

Publication number
WO2013056733A1
WO2013056733A1 PCT/EP2011/068264 EP2011068264W WO2013056733A1 WO 2013056733 A1 WO2013056733 A1 WO 2013056733A1 EP 2011068264 W EP2011068264 W EP 2011068264W WO 2013056733 A1 WO2013056733 A1 WO 2013056733A1
Authority
WO
WIPO (PCT)
Prior art keywords
projection
generating
image
ray radiation
voxel
Prior art date
Application number
PCT/EP2011/068264
Other languages
English (en)
Inventor
Anna Jerebko
Original Assignee
Siemens Aktiengesellschaft
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens Aktiengesellschaft filed Critical Siemens Aktiengesellschaft
Priority to PCT/EP2011/068264 priority Critical patent/WO2013056733A1/fr
Priority to US14/353,116 priority patent/US9449403B2/en
Publication of WO2013056733A1 publication Critical patent/WO2013056733A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/02—Devices for diagnosis sequentially in different planes; Stereoscopic radiation diagnosis
    • A61B6/025—Tomosynthesis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/50—Clinical applications
    • A61B6/502—Clinical applications involving diagnosis of breast, i.e. mammography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/52—Devices using data or image processing specially adapted for radiation diagnosis
    • A61B6/5258—Devices using data or image processing specially adapted for radiation diagnosis involving detection or reduction of artifacts or noise
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—2D [Two Dimensional] image generation
    • G06T11/003—Reconstruction from projections, e.g. tomography
    • G06T11/006—Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00—Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/58—Testing, adjusting or calibrating apparatus or devices for radiation diagnosis
    • A61B6/582—Calibration
    • A61B6/583—Calibration using calibration phantoms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10112—Digital tomosynthesis [DTS]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30068—Mammography; Breast
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2211/00—Image generation
    • G06T2211/40—Computed tomography
    • G06T2211/436—Limited angle

Definitions

  • the present invention relates to the reduction of artifacts in digital breast tomosynthesis and in computer tomography.
  • Another algorithm can involve segmenting out the objects causing the artifacts, such as a needle or other high con- trast objects, and avoiding reconstruction using these pixels, as has been used in computer tomography.
  • Streak artifact removal algorithm similar to those conventionally used in computer tomography technology are often not applicable to tomosynthesis due to the limited acquisition angle.
  • Artifacts in digital breast tomosynthesis are caused by large calcifi ⁇ cations and dense tissue. It is to be understood, that mask- ing-out these anatomical structures prior to reconstruction is not possible, because dense tissue has an amorphous struc ⁇ ture and exact location calcification in the projections is difficult to determine prior to reconstruction.
  • slice thickness filter One of the most efficient ways of out-of-plane artifact sup ⁇ pression is the use of slice thickness filter.
  • the unwanted side effect of this filter is blurring of small anatomical details such as microcalcifications over the total thickness of the slice.
  • This filter does not remove out-of-plane blur caused by dense tissue and large masses, which complicates three-dimensional visualization and reconstruction of thicker slabs . It is an object of the present invention to reduce artifacts in a reconstructed image and/or a projection image.
  • the object of the invention is achieved by a method for re ⁇ constructing and displaying tomography data comprising the step of emitting X-ray radiation through a matter to be analyzed in a plurality of different angles and detecting the X- ray radiation emitted into the matter to be analyzed in a plurality of different projections after the X-ray radiation has passed through the matter with a detector comprising an array of pixels, wherein each pixel generates a grey value based on the received radiation.
  • a volumetric image or a set of projection images is generated based on the grey values acquired by the array of pixels.
  • the step of reconstructing a volumetric image or a set of projection images comprises the step of ignoring or replacing grey values by comparing them with a stored statistical model of image data, wherein the statistical model of the image data is determined by analyz- ing a stored set of image information, e.g. from a plurality of patients and/or a phantom.
  • the object of the invention is also achieved by an apparatus for reconstructing and displaying tomography data comprising an emitter emitting X-ray radiation through a matter to be analyzed in a plurality of different angles and a detector detecting the X-ray radiation emitted into the matter to be analyzed in a plurality of different projections after the X- ray radiation has passed through the matter, wherein the de- tector comprises an array of pixels, wherein each pixel gen ⁇ erates a grey value based on the received radiation.
  • a recon ⁇ structing device is adapted to reconstruct a volumetric image or at least one projection image based on the grey values in the acquired array of pixels by ignoring or replacing grey values by comparing them with a stored statistical model of image data.
  • the apparatus further comprises a model genera ⁇ tion device adapted to generate the statistical model of the image data by analyzing a stored set of image information, e.g. from a plurality of patients and/or a phantom.
  • the apparatus may be configured such as is described below with respect to the method.
  • the statistical model of the image data may be calculated based on a stored set of image information that are sample images generated by emitting X-ray radiation through the mat- ter to be analyzed in a plurality of different angles and de ⁇ tecting the X-ray radiation emitted into the matter to be analyzed in a plurality of different projections after the X- ray radiation has passed through the matter with the detector comprising the array of pixels, wherein each pixel generates a grey value based on the received X-ray radiation.
  • the volu ⁇ metric image and/or at least one projection image is recon ⁇ structed based on the grey values in the acquired array of pixels.
  • a statistical model of the image data may be gener ⁇ ated by analyzing the volumetric image and/or at least one projection images. These steps may be performed by the model generation device.
  • the statistical model of the image data may reflect the sta ⁇ tistical distribution parameters of grey values in acquired projections contributing to the construction of each voxel of a volumetric image or a pixel in reconstructed a projection image.
  • the statistical parameters may comprise a mean value and a standard deviation.
  • the model may comprise a standard deviation for each grey value.
  • the voxel may be calculated based on the sum of all corre ⁇ sponding projection pixels.
  • tomosynthesis reconstruction only a limited amount of projections is available and a large portion of them could be affected by artifacts. Therefore, statistical distribution parameters computed di- rectly from projections pixels could be severally biased.
  • the invention proposes to predict the statistical distribu ⁇ tion parameters based on models, which are derived from ac- quired images.
  • a grey value of a projection pixel is ignored or substituted by a suitable value, if the grey value is does not satisfy the chosen statistical test, wherein the mean or median value is computed from the acquired projection pixel set and all the other distribution parameters are predicted using the precomputed model.
  • a grey value of a projection pixel is ignored or substituted by a suitable val ⁇ ue, if the grey value is not within a pre-configurable range. Therefore, outliers can be removed more reliably.
  • the present invention proposes to calculate the statis ⁇ tical model based on reconstructed images without artifacts which have been calculated beforehand.
  • the model generation device may pre-compute the statistical model of the image data based on intensity distribution pa ⁇ rameters without the presence of artifacts.
  • a data set of real patient im ⁇ ages representing a range of image acquisition parameters, object thicknesses and/or radiographic density may be sam ⁇ pled.
  • the majority of the reconstructed image voxels are ar- tifact free, therefore a fitting a smoothed regression model to these data is equivalent to modeling Artifact-free' dis ⁇ tribution .
  • the model generation device may predict regression-based mod ⁇ el parameters of the distribution of the artifact free pro ⁇ jection values. Thereby, a more accurate statistical model can be provided.
  • the step of generating image information may comprise the step of generating filtered back projections or iteratively reconstructing the image and/or involve a projection-tool- projection reconstruction method.
  • the edge length of a voxel may be set approximately equal to the edge length of a pixel of the detector detecting said X-ray radiation.
  • the length of the edges of a voxel may be set such that the voxel is an isotopic voxel.
  • An isotopic voxel may be a cube.
  • the edge length of a voxel may be set such that for all projection an- gles the voxel projects into an area being smaller than or essentially equal to the size of approximately one pixel of the acquired projection.
  • Max_proj , N 0, in which voxel j is imaged with a minimal amount of overlapping higher intensity voxels corresponding to the dense region intersection with the X-ray Rij .
  • M Max_proj - N, M ⁇ 0, where the projection voxel j overlaps with projections of voxels which correspond to the objects of significantly higher density intersection with the same ray Rij . It is not known a priori if a majority or minority of projec ⁇ tions produces artifacts for a particular voxel due to the non-unimodal distribution of the grey values of the pixel.
  • the Dunn index aims to identify dense and well separated clusters. It is defined as the ratio between the minimal inter-cluster distance to the maximal intra-cluster distance. For each cluster partition, the Dunn index can be calculated based on the inter-cluster distance.
  • a simplified algorithm could be used to reduce the computa ⁇ tion time. For example, a simple approach is to reduce each sample set of projection pixels corresponding to a particular voxel to a unimodal set.
  • a smoothed model of distribution parameters (such as mean and standard deviation) instead of calculated parameters for each mean intensity of a voxel, object thickness, electrode volt ⁇ age, electrode current, volume coordinates (in case of non- uniform noise distribution) may be used.
  • the reason to use the model parameters instead of a calculated parameter is that the calculated distribution parameters would be biased for the voxels where artifacts are caused not by a few pro ⁇ jections (minority) but by a large number of projections.
  • each out- lier is iteratively removed until all data points satisfy the chosen statistical criteria for the view calculated mean.
  • Another approach is to remove or replace all pixel values that do not satisfy the chosen statistical test in one step.
  • This solution provides a faster reconstruction, since it does not involve recalculating the distribution pa ⁇ rameters. In fact, only one distribution parameter is calculated, such as the mean intensity of the pixel.
  • the other pa ⁇ rameters that are more computationally expensive are predicted using the smoothed model which is pre-computed off ⁇ line on a set of training images.
  • the sorting and on-line calculation of distribution parameters is a computationally expensive operation that is used in other artifact-reduction procedures, and such operation is avoided by the present in ⁇ vention .
  • the particular advantage of the present invention is to pre- compute a statistical model of intensity distribution parame ⁇ ters without the presence of artifacts and to avoid on-line computation of these parameters for all voxels. This may be done, as mentioned before, by means of imaging physical phan ⁇ toms containing multiple inserts with attenuation properties of different breast tissues, scanned multiple times with the full range of image acquisition parameters and by variable object thicknesses. Alternatively, a large dataset of real patient images representing the full range of image acquisi ⁇ tion parameters, object thicknesses and densities could be sampled and used as a training set. The statistical model of intensity distribution parameters is then fitted to the data, for example using a regression method.
  • the minimal number of protection pixels remaining in the artifact-free set has to be estimated.
  • the regression-based model parameters of the distri- bution of the artifact-free projection pixel value have to be estimated. For example, distribution parameters for each set of projection pixel values contributing to reconstruction of a particular voxel are predicted based on the mean value of all the corresponding projection pixels, object thicknesses, electrode voltages, electrode currents and/or volume coordi ⁇ nates. Further, the weighting coefficients for statistical outlier detection tests are estimated.
  • the invention suggests the following method for automatic pa- rameter optimization for each particular system.
  • the parameters above could be selected by a human, such as a radiologist on a training set of images, such that all the clinically significant features (micro cal- cifications, masses, mass speculations, Cuper ligaments etc.) are imaged with the highest possible contrast.
  • the parameters are adjusted by maximizing the contrast and edge sharpness in a set of reconstructed clinical features, such as micro calcifications, masses and/or mass speculations.
  • a reference reconstruction method (such as filtered back pro ⁇ jection or an iterative reconstruction with a large number of iterations to ensure visualization of all clinically signifi ⁇ cant features) is used to reconstruct a set of patient im ⁇ ages.
  • a human observer manually outlines or segments clini ⁇ cally significant features in each reconstructed patient im ⁇ age.
  • the system parameters are then chosen to maximize the contrast of the faintest clinical features.
  • An interactive final step could be used for the adjustment of the parameters to the preferences of the human observers.
  • the reconstructing device may be adapted to reconstruct a slice having a thickness of one voxel and to generate a slab that comprises a plurality of slices.
  • the step of generating a slab may comprise generating an average projection, generating an integral projection, generating a maximum intensity projection, generating a linear combination of the average projection and/or integral projection and/or maximum intensity projection perpendicular to the slice plane or coincid ⁇ ing with one or more of the acquired projection angles, gen ⁇ erating a non-linear combination of the average projection and/or integral projection and/or maximum intensity projec- tion perpendicular to the slice plane or coinciding with one or more of the required projection angles.
  • the slabs may be displayed on a display.
  • the radiologist may select a region of the slab. In this region, a slice having a higher resolution than the slab may be shown as a region of interest .
  • a regular filtered back projection method may be modified in order to incorporate the artifact reduction scheme.
  • the present invention suggests to re ⁇ construct slabs consisting of several thin slices with the total thickness S.
  • Each slab contains isotopic voxels with the size equal to the pixel size p of the acquired projec ⁇ tions.
  • the number of thin slices within a slab equals ap- proximately to S/p. The goal is to ensure that each isotopic voxel (cube-like voxel) projects into approximately one pixel of each acquired projection.
  • the present invention suggests for reconstructing each voxel to compute a sum of filtered and bilinear interpolated pixels of all pro ⁇ jections where this voxel was imaged.
  • the present invention suggests using the above artifact re ⁇ duction method applied on each isotopic voxel constituting the slab before computing the sum of the projection values.
  • the projection could be filtered prior to reconstruction, the present invention suggests removing the slice thickness filter or adjusting it to the thickness p, not S.
  • the slab could be col- lapsed into a slice with thickness S.
  • Thicker slices may be generated by performing average intensity projection or a maximum intensity projection of a slab of thin slices in the direction orthogonal to the slice plane (detector plane) .
  • the thick slice could be obtained by projecting each set of thin slices corresponding to one thick slice on a detector plane with the angle of one of the acquired projections.
  • the pro ⁇ jections could be computed using the Siddon method, Joseph method, a distance-driven projector, a separate-footprint projector or any other way of ray casting or forward projecting technique.
  • the full thin-slice volume does not need to be kept in the memory. Only one set of slab slices at a time is required to compute one slice of a thicker slice volume. The resulting slices of thick-slice volume are kept in the mem ⁇ ory, stored on a hard disk or another medium and/or displayed for viewing.
  • the reconstruction device may perform the step of generating a projection image that comprises at least one of the follow ⁇ ing steps: generating a set of forward projections of the re ⁇ constructed digital breast tomosynthesis volume for visualiz ⁇ ing a rotating mammogram, pre-computing forward projections using a set of sequential viewing angels within the order of magnitude of the maximum tomosynthesis angular range, gener ⁇ ating an average projection, generating an integral projection, generating a maximum intensity projection, generating a linear combination of the average projection and/or integral projection and/or maximum intensity projection perpendicular to the slice plane or coinciding with one or more of the ac ⁇ quired projection angles, generating a non-linear combination of the average projection and/or integral projection and/or maximum intensity projection perpendicular to the slice plane or coinciding with one or more of the acquired projection an- gles.
  • a mammogram simulated from digital breast tomosynthesis data could be useful in clinical practice because they keep the look and feel of screening mammographic images, such as full field digital mammography images, while avoiding additional radiation dose needed for acquisition of a real full field digital mammography image in addition to already acquired digital breast tomosynthesis data.
  • a simulated mammogram allows reconstructing originally ac ⁇ quired projections with a higher level of detail, signal to noise ratio and contrast to noise ratio. This is achieved be- cause each projection is reconstructed taken into account in ⁇ formation from all the acquired projections.
  • the projections could be reconstructed to simulate a rotating three- dimensional mammogram by projecting on the detector orthogonal to the ray from the radiation source through the rotation set (different compared to the acquisition geometry) .
  • a se ⁇ quence of projections reconstructed this way forms a three- dimensional rotating mammogram view that efficiently renders three-dimensional information and allows removing or shifting occlusions by rotating the three-dimensional mammogram.
  • Such representation of digital breast tomosynthesis data limits the amount of images to be browsed, transmitted and stored to a chosen number of discrete viewing angles independently of the breast thickness.
  • a three-dimensional rotating mammogram could be si ⁇ mulated from regular reconstructed digital breast tomosynthe ⁇ sis volume, the quality thereof is limited not only by the angular range of the digital breast tomosynthesis itself but also by the reconstructed slice thickness.
  • the reconstructed voxels are highly anisotropic (e.g.
  • the present in ⁇ vention proposes to forward project all isotopic voxels in thin slices of a slab reconstructed with the above isotopic filtered back projection reconstruction with the artifact reduction in the memory.
  • the forward projection could be ray- driven or voxel-driven.
  • the resulting simulated projections should contain the accumulated sum for average intensity pro- jection or maximum intensity projection or a non-linear or linear combination of forward projections of all the isotopic voxels from the slabs.
  • Each pixel value in reconstructed projections corresponds to the ray integral for average intensity projection or maximum over the ray for max ⁇ imum intensity projection or a combination of those of the object intensity function sampled with step size lesser or equal to the projection pixel size.
  • Each sample of the object density function V contains the integrated artifact-free set of projection pixel values Pf corresponding according to the system acquisition geometry to the three-dimensional coordi ⁇ nates of the sample in the object space.
  • the set of original projection pixel values P is obtained from original acquir ⁇ able projections or filtered projections by projecting the three-dimensional voxel sample V onto the detector.
  • the arti ⁇ fact-free set Pf is then obtained from the original set P by applying the artifact reduction scheme described herein.
  • the present invention is also directed to a computer program product comprising code means adapted to execute the steps according to the above described method, when loaded into a processor .
  • Figure 1 shows an exemplary modality used by the present in ⁇ vention
  • Figure 2 shows a schematic diagram of an inventive apparatus
  • Figure 3 shows an exemplary statistical model
  • Figure 4 shows the application of the statistical model when reconstructing a volumetric image or a projection
  • Figure 5 shows an exemplary method for reconstructing a volu- metric image or a projection image applying the statistical model of image data
  • Figure 6 shows a schematic diagram illustrating the genera ⁇ tion of the statistical model of image data
  • Figure 7 shows a prior art image having artifacts
  • Figure 8 shows a coronal slice without artifact reduction
  • Figure 9 shows a coronal slice computed with artifact reduc ⁇ tion
  • Figure 10 shows an axial slab reconstructed without artifact reduction
  • Figure 11 shows an axial slab with artifact reduction
  • Figure 12 shows a maximum intensity projection computed with ⁇ out artifact reduction
  • Figure 13 is a maximum intensity projection with artifact re ⁇ duction .
  • Figure 1 shows an exemplary modality 1 which is used for dig ⁇ ital breast tomography.
  • the modality 1 comprises a radiation source that is arranged on a gantry 10.
  • the radiation source 2 emits a beam 12a, 12b to the detector 4.
  • the breast is ar ⁇ ranged on the detector 4.
  • the breast is compressed by a com ⁇ pression plate 8 that is linked with the detector plate.
  • the gantry 10 and the X-ray radiation source 2 are pivotable within the limited angle range a.
  • X-ray radiation is emitted by the radiation source 2 and the array of pixels of the detector 4 acquire a projection that is output via an amplifier 6. Thereafter, a further projection image is acquired, after the radiation source 2 is positioned under a different angle with respect to the breast and the detector 4.
  • Figure 2 shows an exemplary apparatus 20 for reconstructing and displaying tomography data.
  • a modality 22 generates a plurality of projection images which are transferred to a re- construction device 24.
  • the reconstruction device 24 reconstructs a volumetric image or a set of projection images based on the acquired grey values of the array of pixel by ignoring or replacing grey values by comparing them with a stored statistical model of image data.
  • the volumetric image or the projection image is output on a display device 26.
  • the reconstruction device 24 retrieves the statistical model from a storage unit 30.
  • the apparatus 20 for reconstructing and displaying tomography data further includes a model generation device 28 that gen ⁇ erates a statistical model of the image data by analyzing a stored set of image information from a plurality of patients or a phantom.
  • the statistical model may comprise a standard deviation associated to a mean value.
  • Figure 3 shows a plot of the mean value versus the standard deviation.
  • a mean value of zero corresponds to a standard deviation of approximately 2.5 whereas, a mean value of 225 corresponds to a standard deviation of approxi- mately 25.
  • the plot according to Figure 3 can be determined by the method described above, e.g. by scanning a phantom or by investigating a plurality of patient images.
  • Figure 4 shows a plot of the intensity of different projec ⁇ tion pixels that are assigned to the same voxel. The x axis indicates the grey value, whereas the y axis indicates the number of pixels having approximately the same grey value.
  • Figure 4 is a histogram representation that shows a bimodal curve .
  • the statistical model indicates that the lower grey values represent the expected or true grey value. Therefore, only grey values satisfying statistical model are used by the re ⁇ construction device 24 to reconstruct the image. All other grey values may be ignored or substituted by the mean value.
  • Figure 5 shows a method ac- cording to the present invention.
  • the method starts that step 50.
  • step 52 X-ray radiation is emitted, which passes the body of the patient.
  • step 54 a detector having an array of pixel detects the X-ray radiation. Thereby, the attenuation of the body tissue which the X-ray radiation passed can be determined.
  • step 56 it is determined, whether all neces ⁇ sary projections have been imaged. If not all necessary pro ⁇ jections have been imaged yet, steps 52 and 54 are repeated until all projections are determined, wherein the body tissue is subjected to X-ray radiation under a different angle.
  • step 58 the N projections are stored in a data base.
  • step 60 it is determined whether the grey value GV j is larger or equal than a lower threshold LT and smaller or equal than an upper threshold UT .
  • the lower threshold LT and the upper threshold TH are determined from a statistical mod ⁇ el that has been described with respect to Figures 3 and 4. If the grey value GV j is within the before mentioned allowed range between the lower threshold LT and upper threshold UT, the grey value GV j is used in step 62 for determining the voxel, e.g. by a maximum intensity projection or a mean intensity projection. If the grey value GV j is not within the allowed range, step 64 outputs the mean value of the statistical model as the grey value of the respective pixel of the respective projec ⁇ tion. This method ensures that the allowed mean value is used for determining the voxel value, e.g. by maximum intensity projection or mean intensity projection.
  • Step 66 ensures that steps 60, 62 and 64 are repeated over all M pixels of a projection.
  • Step 68 ensures that steps 60, 62, 64 and 66 are repeated for all N projections. If all M pixels in N projections have been examined and the corre ⁇ sponding voxel value of their volumetric representation calculated is calculated in step 70. Step 70 calculates the volumetric image and/or projection image. The procedure ends in step 72.
  • Steps 152, 154, 156 corresponds to steps 52, 54 and 56, respectively of Fig. 5. For the sake of brevity and clar ⁇ ity the description thereof is not repeated. Steps 152, 154 and 156 are performed over variety of different tissue types and tissue thicknesses. Further, the parameter of the X-ray apparatus, such as electrode voltage, electrode current etc., may be changed wherein one phantom is imaged by a plurality of different parameters of the X-ray apparatus. It is to be understood that instead of a phantom, patients may be used for obtaining the acquired projections for generating the vo ⁇ lumetric images. In step 157 volumetric images and/or projec- tion images are reconstructed from the acquired projections.
  • step 157 volumetric images and/or projec- tion images are reconstructed from the acquired projections.
  • the reconstructed volumetric images and/or projection images are stored in a database in step 158.
  • the database may also be fed with patient data that has been determined to be rele- vant in step 180.
  • step 180 searches a data ⁇ base of existing patient images for relevant images and for ⁇ wards this information to step 158 in which the relevant im ⁇ age data is stored in a database.
  • step 158 is followed by step 182 in which statistical data as shown in Figure 3 is extracted from the volumetric images and projection images in the database.
  • Step 182 may extract sta- tistical distribution parameters such as mean values, stan ⁇ dard deviation and the like as has been described above with respect to the inventive method.
  • the statistical models are stored in step 184 in a database comprising a statistical model of image data. The method ends in step 186.
  • Figure 7 shows an oblique maximum intensity projection visu- alizing out-of-plane-artifacts caused by dense tissue and calcifications in a breast volume reconstructed with a fil ⁇ tered back projection method.
  • Figure 8 shows a one millimeter thick coronal slice computed without artifact reduction using the original filtered back projection algorithm.
  • Figure 9 is an example of a one millimeter thick coronal slice computed with the artifact reduction according to the present invention.
  • Figure 10 shows a two millimeter thick maximum intensity pro ⁇ jection axial slab reconstructed with the original forward back projection without artifact reduction.
  • Figure 11 shows a two millimeter thick maximum intensity pro ⁇ jection axial slab reconstructed with artifact reduction.
  • Figure 12 shows a maximum intensity projection computed from the volume reconstructed with the original filtered back pro ⁇ jection method.
  • Figure 13 shows a maximum intensity projection computed from the volume reconstructed with the artifact reduction method.
  • the present invention reconstructs a set of thin slices with isotropic voxels with a size approximately equal to the pro ⁇ jection pixel size and performs an outlier removal or arti ⁇ fact reduction step on isotropic voxels.
  • the pixel intensity and noise distribution parameters of projection pixel intensity samples are modeled corresponding to each voxel in the object space using a set of training images representing the full range of image acquisition parameters, object thick ⁇ nesses and/or densities.
  • the artifact-free pixel intensity and/or noise distribution parameters are predicted for each voxel in the new test or patient images based on available image acquisition parameters, mean intensities of the projec ⁇ tion pixel sample corresponding to each voxel using the learned model.
  • Predicted artifact-free pixel intensity pa ⁇ rameters and/or noise distribution parameters are used for each voxel in a statistical outlier removal by an artifact reduction algorithm.
  • a weighting factor in a statistical outlier removal test is optimized based on maximizing the contrast-to-noise ratio in a set of clinical features marked in the training images.
  • a statistical test with distribution parameter model is used that is predicted by using the trained model and the weight ⁇ ing factor determined by the above-mentioned method on the thin-slice isotropic voxels and then by computing thicker slices by average or maximum intensity projection of a slab of thin slices in the direction orthogonal to the slice plane, i.e. detector plane or with one of the acquisition angles.
  • the full thin-slice volume does not need to be kept in the memory. Only one set or slab of slices at a time is re- quired to compute one slice of a thicker slice volume.
  • the set of slices belonging to a slab could be kept in the memory or alternatively only one thin slice at a time could be kept in the memory and added to the thick slice incrementally.
  • Slices of thick-slice volume are kept in the memory and/or stored on a hard drive or another medium and/or displayed for viewing .
  • a thick-slice volume is obtained by performing the step above but projecting each set of thin-slices corresponding to one thick slice on the detector plane with the angle of one of the acquired projections.
  • the projections of the digital breast tomosynthesis volume could be computed using a Siddon method, a Joseph method, a distance-driven projector, a separable-footprint projector or any other ray casting or forward projection technique.
  • the present invention provides the advantage that the recon- structed image is de-blurred and streak artifacts are re ⁇ cuted. Further, thick-slice reconstruction or slabbing is possible without the loss of contrast or sharpness. Three- dimensional visualizations and simulated forward projections are possible. An additional efficient noise control through filtering or the use of the prior functions is possible while the boundaries are preserved. Visibility of clinical features is improved and a higher contrast to noise ratio is achieved.
  • the invention proposes a method for out-of-plane artifact re- duction in digital breast tomosynthesis reconstruction.
  • Be ⁇ cause of the limited angular range acquisition in Digital Breast Tomosynthesis the reconstructed slices of prior art methods have reduced resolution in z-direction and are af ⁇ fected by artifacts.
  • the out-of-plane blur caused by dense tissue and large masses complicates three dimensional visu ⁇ alization and reconstruction of thick slices volumes with prior art methods.
  • the streak-like out-of-plane artifacts caused by calcifications and metal clips distort the true shape of calcification that is regarded by many radiologists as an important malignancy predictor.
  • the inventive technique involves reconstructing a set of su- perresolution slices and predicting the Artifact-free' voxel intensity based on the corresponding set of projection pixels using a statistical model learned from a set of training da- ta .

Abstract

La présente invention concerne la réduction d'artefacts sur une tomosynthèse mammaire numérique et une tomodensitométrie. En raison de l'acquisition de plage angulaire limitée en DBT, les coupes reconstruites ont une résolution réduite dans la direction z et sont affectée par des artefacts. Le flou hors plan provoqué par un tissu dense et des masses importantes complique la visualisation en 3D et la reconstruction de volumes de coupes épaisses. Les artefacts hors plan de type stries provoqués par les calcifications et les agrafes métalliques déforment la véritable forme de la calcification qui est considérée par un grand nombre de radiologues comme un important prédicteur de malignités. Les petites caractéristiques cliniques telles que les micro-calcifications peuvent être obscurcies avec des artefacts brillants. La technique implique la reconstruction d'un jeu de coupes en super résolution et la prédiction de l'intensité de voxel « sans artefact » sur la base du jeu correspondant de pixels de projection en utilisant un modèle statistique appris d'un jeu de données d'entraînement. Les images reconstruites obtenues sont défloutées et les artefacts en stries sont réduits, la visibilité des caractéristiques cliniques, le contraste et la netteté sont améliorés, la visualisation 3D et la reconstruction de coupes épaisses est possible sans perte de contraste et de netteté.
PCT/EP2011/068264 2011-10-19 2011-10-19 Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique WO2013056733A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/EP2011/068264 WO2013056733A1 (fr) 2011-10-19 2011-10-19 Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique
US14/353,116 US9449403B2 (en) 2011-10-19 2011-10-19 Out of plane artifact reduction in digital breast tomosynthesis and CT

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/EP2011/068264 WO2013056733A1 (fr) 2011-10-19 2011-10-19 Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique

Publications (1)

Publication Number Publication Date
WO2013056733A1 true WO2013056733A1 (fr) 2013-04-25

Family

ID=44983497

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2011/068264 WO2013056733A1 (fr) 2011-10-19 2011-10-19 Réduction d'artefact hors plan sur une tdm et une tomosynthèse mammaire numérique

Country Status (2)

Country Link
US (1) US9449403B2 (fr)
WO (1) WO2013056733A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11030779B2 (en) 2016-05-24 2021-06-08 Koninklijke Philips N.V. Depth-enhanced tomosynthesis reconstruction
US11399792B2 (en) 2014-09-16 2022-08-02 Dentsply Sirona Inc. Methods, systems, apparatuses, and computer programs for processing tomographic images

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10679104B2 (en) * 2012-10-26 2020-06-09 Varian Medical Systems, Inc. Forward projection cylinder volume method
CN106716177B (zh) * 2014-11-18 2019-04-12 株式会社岛津制作所 断层图像显示装置
US9852526B2 (en) * 2015-04-10 2017-12-26 Toshiba Medical Systems Corporation Method and apparatus of resampling and averaging to obtain tilted thick-slice computed tomography images
CN108369745B (zh) * 2015-12-15 2023-11-07 皇家飞利浦有限公司 条纹伪影预测
US11185294B2 (en) * 2016-02-26 2021-11-30 The Trustees Of The University Of Pennsylvania Super-resolution tomosynthesis imaging systems and methods
US9943280B2 (en) 2016-03-07 2018-04-17 General Electric Company Breast tomosynthesis with flexible compression paddle
JP6707920B2 (ja) * 2016-03-14 2020-06-10 株式会社リコー 画像処理装置、画像処理方法、およびプログラム
US10282918B2 (en) * 2016-09-20 2019-05-07 Siemens Healthcare Gmbh Two-dimensional cinematic medical imaging in color based on deep learning
EP3378401A1 (fr) 2017-03-23 2018-09-26 Siemens Healthcare GmbH Présentation d'une zone d'intérêt
CN107610196B (zh) * 2017-09-04 2020-12-11 东软医疗系统股份有限公司 双能量ct图像处理方法、装置以及图像处理设备
EP3457353B1 (fr) * 2017-09-18 2020-11-25 Siemens Healthcare GmbH Procédé et système pour l'obtention de forme réelle d'objets dans une image médicale
US10679384B2 (en) * 2017-09-29 2020-06-09 General Electric Company Systems and methods for deep learning-based image reconstruction
JP7159341B2 (ja) 2018-01-02 2022-10-24 コーニンクレッカ フィリップス エヌ ヴェ 正則化再構成のための学習ベースのボクセル進化
US11227418B2 (en) 2018-12-28 2022-01-18 General Electric Company Systems and methods for deep learning-based image reconstruction
CN110432919B (zh) * 2019-09-06 2022-11-29 中国人民解放军东部战区总医院 一种基于3d模型的c臂x线片光机实时成像对比方法
WO2021155340A1 (fr) * 2020-01-31 2021-08-05 The General Hospital Corporation Systèmes et procédés de réduction d'artéfacts dans une tomosynthèse avec traitement d'image d'apprentissage profond multi-échelle
DE102021210289A1 (de) * 2020-09-30 2022-03-31 Siemens Healthcare Gmbh Verfahren zum Erzeugen von Ergebnisschichtbildern mit zumindest teilweise verschiedener Schichtdicke

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1998003115A1 (fr) * 1996-07-23 1998-01-29 The General Hospital Corporation Systeme de tomosynthese destine a l'imagerie des seins
US20030194121A1 (en) * 2002-04-15 2003-10-16 General Electric Company Computer aided detection (CAD) for 3D digital mammography
WO2009134970A1 (fr) * 2008-04-30 2009-11-05 Real-Time Tomography, Llc Reconstruction dynamique d’images tomographiques et rendu sur demande
US20110229007A1 (en) * 2010-03-18 2011-09-22 Anna Jerebko Tomosynthesis method with an iterative maximum a posteriori reconstruction

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2851359B1 (fr) * 2003-02-14 2005-05-06 Ge Med Sys Global Tech Co Llc Procede et dispositif de calibration et de correction de niveaux de gris
US7873236B2 (en) 2007-08-28 2011-01-18 General Electric Company Systems, methods and apparatus for consistency-constrained filtered backprojection for out-of-focus artifacts in digital tomosythesis
US8244016B2 (en) 2009-07-20 2012-08-14 Wisconsin Alumni Research Foundation Method for suppressing streak artifacts in images produced with an x-ray imaging system
US8340388B2 (en) * 2009-11-03 2012-12-25 Icad, Inc. Systems, computer-readable media, methods, and medical imaging apparatus for the automated detection of suspicious regions of interest in noise normalized X-ray medical imagery

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1998003115A1 (fr) * 1996-07-23 1998-01-29 The General Hospital Corporation Systeme de tomosynthese destine a l'imagerie des seins
US20030194121A1 (en) * 2002-04-15 2003-10-16 General Electric Company Computer aided detection (CAD) for 3D digital mammography
WO2009134970A1 (fr) * 2008-04-30 2009-11-05 Real-Time Tomography, Llc Reconstruction dynamique d’images tomographiques et rendu sur demande
US20110229007A1 (en) * 2010-03-18 2011-09-22 Anna Jerebko Tomosynthesis method with an iterative maximum a posteriori reconstruction

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11399792B2 (en) 2014-09-16 2022-08-02 Dentsply Sirona Inc. Methods, systems, apparatuses, and computer programs for processing tomographic images
EP3662839B1 (fr) * 2014-09-16 2023-07-19 Sirona Dental, Inc. Procédés, systèmes, appareils et programmes informatiques pour le traitement d'images tomographiques
US11030779B2 (en) 2016-05-24 2021-06-08 Koninklijke Philips N.V. Depth-enhanced tomosynthesis reconstruction

Also Published As

Publication number Publication date
US9449403B2 (en) 2016-09-20
US20140294138A1 (en) 2014-10-02

Similar Documents

Publication Publication Date Title
US9449403B2 (en) Out of plane artifact reduction in digital breast tomosynthesis and CT
US9451924B2 (en) Single screen multi-modality imaging displays
Sechopoulos A review of breast tomosynthesis. Part II. Image reconstruction, processing and analysis, and advanced applications
US7676257B2 (en) Method and apparatus for segmenting structure in CT angiography
US9401019B2 (en) Imaging tomosynthesis system, in particular mammography system
EP2410491B1 (fr) Système et procédé de reconstruction d'images à rayons X
US8280135B2 (en) System and method for highly attenuating material artifact reduction in x-ray computed tomography
US8229200B2 (en) Methods and systems for monitoring tumor burden
US11660061B2 (en) Method and system for motion assessment and correction in digital breast tomosynthesis
US11816764B2 (en) Partial volume correction in multi-modality emission tomography
JP2016522071A (ja) トモシンセシスデータからエッジ保存合成マンモグラムを生成するための方法
US9047498B2 (en) Method for reconstructing a reconstruction data set containing two-dimensional virtual X-ray images
von Falck et al. Informatics in radiology: sliding-thin-slab averaging for improved depiction of low-contrast lesions with radiation dose savings at thin-section CT
EP3378041B1 (fr) Reconstruction d'image de pet et traitement au moyen de substitutions de lésion
EP1716537B1 (fr) Dispositif et procede de traitement de d'images en coupe
US9349199B2 (en) System and method for generating image window view settings
US11580678B2 (en) Systems and methods for interpolation with resolution preservation
KR102493193B1 (ko) 다중 수집을 통해 콘트라스트를 정규화하는 방법 및 시스템
JP2022547463A (ja) コーンビームｃｔにおけるニューラルネットワークに基づく限定角度アーチファクト軽減のための信頼度マップ
WO2017216248A1 (fr) Détermination de la teneur en calcium à partir de données de tomodensitométrie spectrale
US10755454B2 (en) Clinical task-based processing of images
CN108366772B (zh) 图像处理装置和方法
Passand Quality assessment of clinical thorax CT images
Cowart Task-based assessment of digital breast tomosynthesis: effect of anatomy from multiple anthropomorphic 3D printed phantoms
Andersson Effects of slabbing and slice thickness on image reading time and micro calcification detection in breast tomosynthesis

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11771147

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 14353116

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11771147

Country of ref document: EP

Kind code of ref document: A1